Detectron2: Error when evaluate: object of type <class 'numpy.float64'> cannot be safely interpreted as an integer

[hungthanhpham94]

Instructions To Reproduce the Issue

1. what changes you made (git diff) or what code you wrote
   I don't change code in Detectron2 repo
   I just prepare COCO dataset following instruction
   git diff is empty

2. what exact command you run:
   Training: python tools/train_net.py --num-gpus 2 --config-file configs/COCO-Detection/retinanet_R_50_FPN_3x.yaml SOLVER.IMS_PER_BATCH 4 SOLVER.BASE_LR 0.0025
   Evaluate: python tools/train_net.py --config-file configs/COCO-Detection/retinanet_R_50_FPN_3x.yaml --eval-only MODEL.WEIGHTS output/model_final.pth

3. what you observed (including the full logs):
   

4. please also simplify the steps as much as possible so they do not require additional resources to
   run, such as a private dataset.

Expected behavior

I expect that process will show AP result

Environment

Thanks for your help!

[ppwwyyxx]

It's an issue of coco API: it does not work with numpy 1.18. Downgrading to 1.17 will help.

I'll try to fix this in COCO API.